4. SOLID-PHASE EXTRACTION–GAS
4.1. Solid-Phase Extraction–Gas Chromatography–Mass
Next to trace-level detection, unambiguous confirmation of the presence of target compounds and the provisional identification of unknowns is rapidly gaining im- portance. Some five years ago, several papers were published that demonstrated that SPE–GC–MS can do just that. In an early study [Ref. i of Table 6], atrazine and simazine were determined by means of SPE–GC–MS, in both the multiple- ion-detection (MID) and full-scan (FS) modes, as well as SPE–GC–NPD, using 1- and 10-ml samples. Quantification of the microcontaminants at levels of ap- proximately 10 to 50 ng/L presented no real problems with relative standard deviations (RSDs) of 3 to 8% (n⫽ 4) and limits of detection as low as 0.5 ng/
L (MID), 3 ng/L (FS) and 4 ng/L (NPD) for 10-ml samples. The MS- and NPD- based data showed good agreement (with differences generally being less than 10 ng/L at the 10 to 70 ng/L level) and linear calibration curves were obtained in both instances.
Attention was also devoted to nontarget analysis. A mere 1-ml sample of Rhine River water was spiked with 1àg/L of each of a mix of 168 microcontami-
180 Hankemeier and Brinkman
nants, and a 1-minute window selected for further study [Ref. i of Table 6]. By selecting a number of individual masses found in the mass spectra of the apex of each of the six major peaks, it was possible to record rather undisturbed mass traces such as those shown in Figure 8. From these traces, it was apparent that at least nine peaks eluted in the selected time window. The mass spectra of the nine peaks were recorded at their apexes and compared with the NBS library.
The data of Table 5 show that the final result was quite satisfactory. Problems started to occur when the peak maxima were merely 0.01 to 0.02 minute apart, as is demonstrated by the pair 1-nitronaphthalene/3-chloro-4-nitrophenol.
In subsequent studies, the development and use of an automated benchtop instrument received much attention. The final system (see Fig. 3) consisted of a Prospekt (Spark Holland, Emmen, Netherlands) automated sample-handling module for trace enrichment, drying of the SPE cartridge, and analyte transfer under PCSE conditions using an SVE and an on-column injector, coupled on-
Figure 8 (a) Nine reconstructed ion traces (masses indicated above each peak) of on- line SPE–GC–MS of 1 ml of Rhine River water spiked at the 1àg/L level with each of 168 micropollutants and (b) mass spectra of phenylacetic acid and 1,4-dibutoxybenzene, which elute at 21.83 and 21.86 min, respectively. (Ref. i of Table 6.)
On-Line Sample Preparation for Water Analysis 181
Table 5 Compounds Identified in the Time Window of 21.4–22.4 Minutes after SPE–GC–MS of 1 ml of River Water Spiked at 1àg/l
Retention Main mass Library
time (min) (m/z) Compound search fita
21.50 149 Diethyl phthalate 0.94
21.58 119 N,N-Diethyl-3-methylbenzamide 0.91
21.80 284 Hexachlorobenzene 0.85
21.83 91 Phenylacetic acid 0.95
21.87 110 1,4-Dibutoxybenzene 0.96
21.93 182 Benzophenone 0.95
22.17 173 1-Nitronaphthalene 0.66
22.18 99 3-Chloro-4-nitrophenol 0.61
22.28 106 3-Aniliniopropionitril 0.95
aLibrary search fit factor on a scale of 1.00.
Source: Ref. i of Table 6.
line to a GC–MS. The total system was completely software-controlled under Microsoft Windows and was used to analyze a variety of water samples. The further development and subsequent upgrading of that setup, with a self-con- trolled system as the ultimate goal, was discussed previously in section 3.2. The discussions below, therefore, mainly emphasize the detection/identification per- formance that can be achieved when SPE is on-line combined with GC.
Table 6 summarizes relevant analytical data on the sensitivity of SPE–GC–
MS of, chiefly, aqueous environmental samples such as tap water, surface water, and wastewater. The main conclusions that can be drawn from the data are rather promising. Sample volumes of about 10 ml suffice to obtain full-scan MS traces such as are shown in Figure 9. Detection limits were in the 20 to 50 ng/L range or lower for essentially all compounds. As a demonstration of the identification power of the procedure, the traces of the four characteristic ions of peak 11 (benz- aldehyde) in the raw, i.e., nonspiked, water are included. Comparison with the 0.5-àg/L spiked trace shows that benzaldehyde was present at a level of approxi- mately 40 ng/L. This system is well suited for the screening of rather volatile as well as high(er)-boiling compounds.
While monitoring studies often aim at detecting ‘‘all’’ microcontaminants present above a threshold level (estimated from, e.g., FID or a total-ion-current trace), there are also situations in which targeted analysis is the main goal. Then, it is beneficial to use (time-scheduled) selective ion monitoring (SIM) and related techniques. Figure 10 shows a relevant example. The detectability of the com- pounds of interest improved 3- to 10-fold upon going from the total ion chromato- gram to post-run ion extraction, and improved a further 10-fold upon going from
182HankemeierandBrinkman
Table 6 Selected Applications of On-Line SPE–GC with MS or Tandem MS Detection for River and Tap Water Sample
Analytes MS detection volume (ml) LOD (ng/l) Ref.
Atrazine, simazine, various SIM 10 0.5 i
micropollutants Full-scan 3
Various micropollutants Full-scan 10 2–50 37, 41, 52, 67, 87, 88, ii
Pesticides, phenols SIM 10 1–20 55, 66
Chlorinated pesticides EI full-scan 100 1–30 89
NCI full-scan 0.1–3
Pesticides MS/MS 10 0.01–2 53
Hetero-atom containing AED/full-scan 5–50 1–15 95, 96
pesticides/micropollutants MSa
aAt-line setup because of necessary repeated injections for AED analysis.
Abbreviations: SIM, selected ion monitoring; NCI, negative chemical ionization; if not stated otherwise, EI ionization was applied.
i.) A.-J. Bulterman, J.J. Vreuls, R.T. Ghijsen and U.A.Th. Brinkman, J. High Resolut. Chromatogr., 16 (1993) 397.
ii.) A.J.H. Louter et al., Intern. J. Environ. Anal. Chem., 56 (1994) 49.
On-Line Sample Preparation for Water Analysis 183
Figure 9 Total ion chromatogram for SPE–GC–MS of 10 ml of Rhine River water (B) nonspiked and (A) spiked at the 0.5àg/L level with 86 microcontaminants. 50àl of methyl acetate were used as presolvent. The insert (C) shows the extracted ion chromatograms of four characteristic masses of benzaldehyde (m/z 51, 77, 105 and 106). (From Ref. 41.)
full-scan acquisition to SIM detection (two ions per analyte) [67]. Detection limits of 0.2 to 1.1 ng/L were achieved for 10-ml surface water samples. However, one should always consider that the improved selectivity and detectability are accompanied by a serious loss of information on the general composition of the sample.
Jahr [55] used Autoloop–GC–MS for the trace analysis of phenols in water at the low-ng/L level. The phenols were derivatized by in-sample acetylation with acetic acid anhydride prior to automated SPE–GC–MS. The method was validated with 26 alkyl- , chloro- , and mononitrophenols; these included 4-nonyl- phenol and 17-ethinylestradiol. Repeatability was good and the sensitivity in the time-scheduled SIM mode was excellent.
On-line dialysis–SPE–GC–MS was developed for the determination of benzodiazepines in plasma [85]. Clean-up was achieved by dialysis of 100-àl samples for 7 minutes using water as the acceptor and trapping the diffused ana- lytes on an SPE column. After drying, the analytes were desorbed with 375àl of ethyl acetate on-line to the GC–MS via a loop-type interface. Sample clean- up was very efficient and offered the possibility of adding chemical agents that
184 Hankemeier and Brinkman
Figure 10 SPE–GC–MS chromatograms of 10 ml of Rhine River water spiked at the 0.1àg/L level. (a) Full-scan mode (m/z 50–375) and (b) time-scheduled MID. Peak as- signment and ions used: 1, mevinphos (m/z 127/192); 2, diazinon (m/z 197/204); 3, feni- trothion (m/z 277/260); 4, fenthion (not determined with MID); 5, triazophos (m/z 161/
257); 6, coumaphos (226/362). (From Ref. 67.)
On-Line Sample Preparation for Water Analysis 185
can help to reduce drug–protein binding. The benzodiazepines were determined in plasma at the 1 ng/ml level which is relevant for forensic or pharmacokinetic studies.